why the neurons are not -1 or 1 after BinaryConv layer?

[Viscoo]

I have run the code successful, but when I check the neurons after the binaryconv layer，but it distribute dispersedly，and have no trend to -1 or 1。

[yaysummeriscoming]

The activations are binarised at the input to the next layer.

as for the final layer: my understanding is that whilst the convolution multiplications are binary (xnor operation), the summing is not:

y = W*X + B y = xnor(w1, x1) + xnor(w2, X2) ... + B

[yaysummeriscoming]

FYI theres another really good paper that has come out that discusses this a bit further and also has some big improvement on the original BinaryNet implemented here. It's a pretty small extension to BinaryNet, hopefully I'll get some time to do an implementation:

How to Train a Compact Binary Neural Network with High Accuracy? https://www.aaai.org/ocs/index.php/AAAI/AAAI17/paper/download/14619/14454

[wjtan99]

@yaysummeriscoming the results in this paper is impressive. When will you implement it? Thanks.